Skip to content
BY 4.0 license Open Access Published by De Gruyter Open Access July 19, 2023

Irrealis-marked interrogatives as rhetorical questions

  • Colin Brown EMAIL logo
From the journal Open Linguistics


I describe and compare two strategies to form rhetorical questions (RQs) in Sm’algyax (Tsimshianic). I show that one kind is isomorphic to ordinary, information-seeking questions, and is compatible with positive and negative answers, while the second is marked with irrealis morphology and only allows negative answers. I provide evidence from answerability and embedding to suggest that both types of RQs in Sm’algyax behave like questions in terms of their syntax/semantics, and propose that the irrealis subordinator present in the second type signals to the addressee that a negative answer is expected. These findings have implications for the presence of irrealis and subjunctive morphology appearing in RQs crosslinguistically.

1 Introduction

Rhetorical questions (RQs) may informally be characterized as interrogative sentences that have the feel of an assertion, that do not expect an answer, but that may be answered (Biezma and Rawlins 2017, Caponigro and Sprouse 2007, Han 2002, Sadock 1974, a.o.). In this article, I discuss RQs in Sm’algyax, also known as Coast Tsimshian, a Maritime Tsimshianic language of British Columbia and Alaska, and show that there are two ways to form them. The first construction is morphosyntactically isomorphic to information-seeking questions (ISQs). In the case of content (or wh-) questions, this involves the appearance of a wh-expression appearing in the initial position, as well as the optional appearance of the interrogative clitic u (1).

(1) Naa(yu) int ba’an gip’aayk’nsk?
naa(=u) in=t baa-’n gip’aayk’nsk
who(=q) ax=3.i run-caus plane
‘Who can fly a plane?’ ISQ/RQ

The second type also features an initial wh-expression, but is further marked by the presence of the irrealis subordinating element dzi, as well as the obligatory absence of the interrogative clitic u.

(2) Naa dzi int ba’an gip’aayk’nsk?
naa(*=u) dzi in=t baa-’n gip’aayk’nsk
who(*=q) irr ax=3.i run-caus plane
‘Who can fly a plane?’ = nobody can fly a plane RQ

Beyond the morphological differences between these constructions, we also observe interpretive differences. The first type in (1) is compatible with positive and negative answers, while the second type is only compatible with negative answers, as indicated by the negative declarative paraphrase given as an alternate translation in (2).

This article presents a description of RQs in Sm’algyax, focusing on the notions of answerability, embeddability, and their appropriateness across different contexts. I investigate why a question marked with dzi would yield an obligatorily rhetorical interpretation, and how this effect could be derived compositionally. I take as a baseline the claim in Caponigro and Sprouse (2007) that RQs share a syntax and semantics with ordinary, ISQs, differing only in terms of their pragmatics, and suggest that unmarked RQs such as (1) are straightforwardly accounted for by this analysis. However, irrealis-marked RQs are obligatorily rhetorical, and have an obligatory “negative implication” distinguishing them from ISQs and unmarked RQs. Drawing from an analysis in Ladusaw (1980) in which questions are worded in a way that facilitates the answer, I suggest that this negative implication arises due to the presence of the irrealis morphology in the question limiting the possible answers in the answer set to the negative answer. This proposal has implications for the analysis of the presence of irrealis or subjunctive marking in interrogative sentences in languages beyond Sm’algyax.

This article proceeds as follows. Section 2 introduces the data and methodology. Section 3 provides a background to canonical vs non-canonical questions and describes canonical question formation in Sm’algyax and the irrealis subordinator dzi. Section 4 covers RQs in Sm’algyax and outlines the differences between the unmarked RQs and the irrealis-marked RQs. In Section 5, I outline an analysis of RQs from Caponigro and Sprouse (2007) for both types of RQs in Sm’algyax but suggest that the negative implication associated with marked RQs arises due to the presence of irrealis morphology coercing a negative answer (compatible with Ladusaw 1980). Section 6 concludes.

2 Data and methodology

All uncited Sm’algyax examples are from my own fieldwork, carried out in Prince Rupert, British Columbia, between 2018 and 2023 with fluent Sm’algyax speakers: Velna Nelson, Ellen Mason, (Txałgiiw/Hartley Bay), and Beatrice Robinson (Gitxaała/Kitkatla), and on Zoom between 2020 and 2023 with Velna Nelson. The elicitations in Prince Rupert most often consisted of a group of two or more language consultants and myself, while Zoom elicitations were one-on-one. Though the data set is not publicly available at the time of publication, data inquiries may be addressed to the author’s email.

All language data are elicited in line with standard semantic fieldwork methodology (Matthewson 2004): the consultant is provided with a context and an English sentence and is asked to translate the sentence into Sm’algyax. Acceptability judgments are elicited by providing the consultant with a Sm’algyax sentence and asking for comments on acceptability for a given context. These examples are presented here, when relevant, alongside the context that was provided during elicitation. Given that this article discusses indirect speech acts (in this case, interrogative sentences being used in conversation in functions distinct from “asking”), some of my elicited examples include constructed dialogues. To construct these dialogues, individual sentences were elicited in context, then assembled, and presented to consultants as a dialogue, which was judged by consultants for acceptability/felicity.

The four-line glossing convention used here is as follows: the top line appears in the community orthography used throughout British Columbia, adapted from John Dunn’s Sm’algyax orthography (Dunn 1979). The second line utilizes the same orthography, but indicates morpheme boundaries; the third line provides grammatical category labels in line with the Leipzig glossing rules,[1] and the final line provides an English translation. A number of examples are pulled from the traditional Ts’msyen narratives (or adaawx), including from Boas (1912) and Beynon (1932–1939), where irrealis-marked RQs commonly occur. Though it was not in use at the time of writing/transcription, I have presented these data in the community orthography.

3 (Non-)canonical questions

Recent work (Farkas 2022) defines a canonical question act as one that satisfies all of the assumptions in (3):

(3) Default assumptions accompanying question acts (Farkas 2022)
a. Speaker ignorance: The speaker’s epistemic state is neutral relative to the possible resolutions of the issue she raises.
b. Addressee competence: The speaker assumes that the addressee knows the information that settles the issue she raises.
c. Addressee compliance: The speaker assumes that the addressee will provide this information in the immediate future of the conversation as a result of the speaker’s speech act.
d. Issue resolution goal: It is assumed that the main aim the speaker pursues when raising an issue is to have it resolved in the immediate future of the conversation.

More informally, a canonical question is one in which the speaker doesn’t know the answer, expects the addressee to know the answer, and expects the addressee will answer the question (as having the question answered is the purpose of asking it in the first place).

However, according to Farkas (2022), one or more of these default assumptions in (3) may be suspended or weakened. This results in a non-canonical question. RQs, a subtype of non-canonical question, arise by suspending the speaker ignorance and issue resolution goal assumptions; the speaker already knows the answer (by the suspension of speaker ignorance), assumes that the addressee also knows the answer (by retaining addressee competence), and does not expect the addressee to answer (by the suspension of the issue resolution goal). This configuration of assumptions results in a sentence that may have an interrogative form (in terms of its morphosyntax, and potentially even semantics), but has a distinct conversational function, more in line with the notions of “asserting,” or “reminding,” rather than “asking.”

3.1 Questions in Sm’algyax

Sm’algyax is an ergative, head-marking language with a rigid verb–subject–object order. Before we turn to our description of RQs in Sm’algyax, let us first outline how this language forms questions. I refer the reader to Forbes (2023) and references within for an overview of the basic grammatical properties of Sm’algyax and its Tsimshianic relatives, and to Brown (2022) for a more detailed description of questions in Sm’algyax.

Polar questions, or yes–no questions, (shown in (4)) retain the default word order (shown in (4)), and are marked by the presence of the clitic ii, which appears exclusively in polar and alternative interrogatives.

(4) (a) Dawłit Dzeen
dawł=t Dzeen
leave=pn Jane
‘Jane left.’ Baseline
(b) Dawłiit Dzeen?
dawł=ii=t Dzeen
leave=q=pn Jane
‘Did Jane leave?’ Polar question
(c) dawłdiit Dzeen?
a=ł dawł-t=ii=t Dzeen leave-3.ii=q=pn Jane
‘Did Jane leave?’[2] Polar question

In the following, we see that information-seeking polar questions (5) and rhetorical polar questions (6) are morphosyntactically identical. The question in the context of (5) clearly satisfies the default assumptions in (3): the speaker doesn’t know the answer and expects their addressees to know and to provide the answer. In (6), the speaker has an answer in mind (namely “no”) and assumes that the addressee shares this answer – (6) may felicitously go unanswered.

(5) ISQ context (from Braun et al. 2019): At a party, you offer cake made with limes. You would like to know which of the guests like this fruit and would like some of it. You say to your guests:
Anoogas ligit naayii lime?
anoox-i[-t]=s ligi=t naa=ii=a lime
like-tr-3.ii=pn ligi=pn who=q=cn lime
‘Does anyone like lime?’
(6) RQ context (from Braun et al. 2019): Your aunt offers limes to her guests. However, it is known that this fruit is too sour to be eaten on its own. You say to your cousin:
Anoogas ligit naayii lime?
anoox-i[-t]=s ligi=t naa=ii=a lime
like-tr-3.ii=pn ligi=pn who=q=cn lime
‘Does anyone like lime?’

As rhetorical polar questions do not have a comparable irrealis-marked construction, the remainder of this article is dedicated to discussing content questions.

Content questions, also known as wh-questions, are characterized by a wh-expression appearing in the initial position, and extraction morphology appearing either before or after the predicate, depending on the grammatical role of the extracted element. A distinct interrogative clitic u appears in the root-level content questions.

(7) (a) Tgi k’apaaytga ’yuuta
tgi k’apaaytk=a’ yuuta
down fall=cn man
‘The man fell down.’ Baseline
(b) Naayu tgi k’apaaytgit?
naa=u tgi k’apaaytk-it
who=q down fall-sx
‘Who fell down?’ Content question

This interrogative particle may not appear in embedded questions.

(8) Güüdaga’nut Klalens goo dm hasagu.
güüdax-’nu=t Klalens [goo(*=u) dm hasax-u]
ask-1sg.iii=pn Clarence [what(*=q) prosp want-1sg.ii]
‘Clarence asked me what I want.’

Interrogative particles appear in ordinary ISQs (7) as well as non-canonical questions such as self-directed questions (those in which the speaker = the addressee (9)) and quiz-show questions (those in which the speaker knows the answer and is testing the addressee (10)). They appear in the first type of RQ (11), but not the second type, which is marked with dzi (12). Though the contexts in which (9) though (11) occur show that they are non-canonical, they are morphosyntactically isomorphic to ISQs.

(9) Context: You’re home alone and you can’t find your keys. You ask yourself:
Ndeyu nahak’a’ayu?
ndeh=u na-hak’a’a-u
where=q poss-key-1sg.ii
‘Where are my keys?’ Self-addressed question
(10) Context: A geography teacher is quizzing a student who is looking at a map of Canada:
Ndeyu wil t’aa ts’a’mas?
ndeh=u wil t’aa=a ts’a’mas
where=q comp sit=cn Vancouver
‘Where is Vancouver?’ Quiz-show question
(11) Context: You’re visiting a friend in a small town and realize you forgot your toothbrush, it’s Sunday night, and you know everything will be closed:
(Oo du,) Ndeyu dm gooyu gya’wn
(oo du) ndeh=u dm goo-u gya’wn
(oh particle) where=q prosp go-1sg.ii now
‘Oh, where could I go now.’ = there’s nowhere to go now Unmarked RQ
(12) Context: You are trying to console a friend who is having a hard time, but you are lost for words:
Goo dza hawi?
goo(*=u) dzi haw-i
what(*=q) irr say-irr.1sg.ii
‘What could I say?’=there’s nothing I can say Dzi-marked RQ

In sum, content questions involve a wh-expression appearing in initial position as well as the optional presence of the interrogative clitic u. The interrogative clitic is restricted to root interrogative sentences and may appear in several non-canonical question types. However, it may not appear in irrealis-marked RQs, which are instead marked with dzi. Let us now outline the distribution of dzi.

3.2 Irrealis marking in embedded and root clauses

The element dzi, which is variably spelled/pronounced as dza, is described as a “subordinator used to express hypotheses” in Sasama (2001) and as a particle that “weakens a statement” in Mulder (1994). Descriptive and theoretical work on the interior Tsimshianic cognate ji/ja give it the label “irrealis” (Rigsby 1986, Tarpent 1987, a.o.), and is described in Hunt (1993) as a subordinator “used when the subordinate clause expresses uncertain or non-factual information.” Though I do not provide an analysis of dzi here, I gloss it as irrealis in this article. Besides appearing in marked-RQs, dzi also appears in a number of environments, exhibiting behavior consistent with a clausal subordinator or complementizer.

One common environment in which dzi appears is embedded polar questions such as (13).[3]

(13) Güüdagat dzit k’otsdit Lucy hoon
güüdax-i-t [dzi=t k’ots-t=t Lucy=a hoon]
ask-tr-3.ii [irr=3.ii cut-3.ii=pn Lucy=cn fish]
‘He asked if Lucy cut the fish.’

Certain non-interrogative clausal complements are also introduced by dzi. In (14b), we see a dzi-clause selected by the predicate “think,” in a context in which the speaker likely no longer believes the embedded clause (that the addressee possesses a dog) to be true.

(14) Context (elicited via storyboard “Feeding Fluffy” TFS Working Group): Pat is tasked with taking care of Fluffy, Stacy’s pet. Pat doesn’t know what kind of animal Fluffy is and buys a bone just in case Fluffy is a dog (Fluffy is in fact a snake). Stacy asks (14a), and Pat responds (14b):
a. Goł mi gan giikdu sayp?
go=ł mi gan giik-t=u sayp 2.i reas buy-3.ii=q bone
‘Why did you buy a bone?’
b. Ha’ligoodi dzi haasgn
ha’ligoot-i [dzi haas-k-n]
think-irr.1 sg.ii [irr dog-pass-2 sg.ii]
‘I thought you had a dog.’

In the following, we see a contrast between embedded clauses introduced by the irrealis subordinator dzi and the default complementizer wil. In (15), dzi introduces a clause that does not appear to be presupposed by the speaker and encodes the otherwise factive predicate aam ‘be good’ with desiderative meaning, while in (16) the embedded clause does appear to be presupposed and there is no special desiderative meaning encoded.

(15) Aam dza aamt.
aam [dza aam-t]
good [irr good-3.ii]
‘I hope that she is good.’ (SLLTD) Literally: It’s good if s/he is good.
(16) Aam wilt siip’ndit Billt Meeli
aam [wil=t siip-’n-t=t Bill=t Meeli]
good [comp=3.i sick-caus-3.ii=pn Bill=pn Mary]
‘It’s good that Bill loves Mary.’

A similar contrast is observed in the following examples. In (18), the speaker does not know if the embedded clause is true, and dzi appears, while in (17), the speaker does know that the embedded clause is true, and wil appears.

(17) Context: You don’t know whether Lucy cut the fish. You’re talking about whether Bill saw Lucy cutting the fish:
Akadit nii dzit k’otsdit Lucy hoon
aka=di=t nii[-t] [dzi=t k’ots-t=t Lucy=a hoon]
NEG=FOC=3.I see[-3.II] [IRR=3.I cut-3.II=PN Lucy=CN fish]
‘He didn’t see whether Lucy cut the fish.’
(18) Context: You don’t know whether Lucy cut the fish. You’re talking about whether Bill saw Lucy cutting the fish:
Akadit nii wilt k’otsdit Lucy hoon.
aka=di=t nii[-t] [wil=t k’ots-t=t Lucy=a hoon]
NEG=FOC=3.I see[-3.II] [COMP=3.I cut-3.II=PN Lucy=CN fish]
‘He didn’t see that Lucy cut the fish.’

In non-clearly embedded contexts, dzi appears in certain polite imperative constructions, as follows:

(19) Dzi amaniidzn dziłam baała wan
dzi amaniist-n dzi=ła=m baał=a wan
irr careful-2sg.ii irr=incep=2sg.i cut=cn deer
‘Take care when you cut open the deer.’ (SLLTD)

Although the clause headed by dzi in (19) is not clearly embedded (that is, it does not appear as a complement to a clear-cut matrix predicate such as those in (13)–(15)), imperatives in Sm’algyax bear morphosyntactic marking indicative of dependent clauses – for instance, the suffixal agreement occurring on the predicate amaniist in (19) indexes agreement with the intransitive subject, which only occurs in dependent clauses (in independent clauses this suffix only indexes agreement with transitive subjects, see Sasama (2001, 146)). This suggests that perhaps (19) is an example of root-level embedded phenomena or “insubordination” (Evans 2007), and that dzi functions uniformly as a subordinator in Sm’algyax. The polite imperative in (19) would then closely resemble imperatives headed by subordinators and complementizers in languages such as French and Polish.

(20) Si on allait se promen-er?
if one went refl walk-inf
‘What if we went for a walk?’ (Evans 2007, French)
(21) Żeby ciocia teraz może zadzwoni-ła
comp auntie now perhaps telephone-pst,f
‘If you (auntie) could perhaps make a phone call for me?’ (Evans 2007, Polish)

4 RQs two ways

This section compares what I refer to as unmarked RQs and irrealis-marked RQs. I show that the unmarked flavor is answerable and compatible with positive and negative answers. Turning to the irrealis-marked flavor, I show that their distribution is more limited, and that they appear in contexts in which the speaker believes that there is no positive answer to the question. In this way, they may function similarly to, and are often translated as, negative declarative sentences. Despite these differences, evidence from embedding suggests that irrealis-marked RQs share syntactic and/or semantic properties with canonical questions.

4.1 Unmarked RQs

This section briefly describes what I refer to as “unmarked RQs.” I refer to them as such due to their morphosyntactic isomorphism to canonical content questions. Unmarked RQs are compatible with negative and positive answers. In (22), the context shows that the speaker (Ben) already knows the answer to the question, and it is negative: “nobody,” while in (23), we see the context shows that the speaker (Ben) knows that the answer is positive: “Cassie.”

(22) Context: Ben is bitter because nobody at his job helped him when he needed it. Later he was invited to a work party. Ben says:
Akadi hasag dm yaayi Awil naayu int łimoomu ła
aka=dii hasax=ł dm yaa-i awil naa=u in=t łimoom-u ła
neg=foc prosp walk-irr.1sg.ii because who=Q ax=3.i help-1sg.ii prox
hasagu a łimoo’ma?
hasax-u a łimoom-’a
want-1sg.ii prep help-pass
‘I don’t want to go. After all, who helped me when I needed help?’ “Negative” RQ
(23) Context (adapted from Caponigro and Sprouse 2007): Al is worried Cassie didn’t have fun at his party. They both know that Cassie was dancing all night at the party. Ben says to Al “Gooyu nahawn,” [what do you mean,]:
Naayu nah miilgit a ludaba aatk?
naa=u nah miilk-it a ludaba aatk?
who=q pfv dance-sx prep whole night
‘Who was dancing all night?’ “Positive” RQ

Though they may felicitously go unanswered, unmarked RQs may be answered. Responses include answers to the question (typical of a question act), agreement or disagreement (typical of an assertion), or silence (typical of assertions, some non-canonical questions).

(24) Context = Al is responding to Ben’s RQ in (23):
A: Cassie…/Oo sm hen [Yes, you’re right]/*silence*

The rhetorical content questions described here are as follows. They are content questions marked with the interrogative clitic u and a wh-expression which undergoes movement (identical to ISQs). They need not be answered but may be answered. They are compatible with positive or negative answers.

I assume that ISQs and unmarked RQs share a basic syntactic structure. In content questions, the complementizer phrase (CP) projection is headed by an interrogative C-element, which in matrix questions optionally surfaces as u, the question particle, and the wh-expression moves from its in-situ position to the specifier position of CP.

(25) ISQs and unmarked RQs

4.2 Marked RQs

“Marked RQs,” in contrast to ISQs and unmarked RQs, are marked with the irrealis subordinator dzi and lack the interrogative clitic u. This construction occurs frequently in narrative contexts and is variably translated as a content question or as a declarative sentence with a negative existential element such as “nowhere,” or “nothing.” Marked RQs are always associated with what I refer to as a “negative implication”: the implication that the speaker believes there is no answer to the question. In these narrative contexts, they are never answered.

In the following, we see examples of marked RQs occurring in texts, with the narrative context indicating that there is no clear answer to the question. In (26a), the narrator utters the marked RQ “Therefore what then could he use now?” after making it clear that Asdiwaal has nothing to use to get himself out of the situation he is in, while in (26b) we see the marked RQ “where could he go?” preceding expository material asserting that there is in fact nowhere to go.

(26) Narrative context: Asdiwaal carries with him a number of magical tools that have gotten him out of tricky situations. However, this time he is stranded on a mountain in a storm without his magical tools:
a. Gan goo dzi gyik hoyt gya’wn?
gan goo dzi gyik hoy-t gya’wn
reas what irr again use-3.ii now
‘Therefore what then could he use now?’
b. Ndaa dzi yaakit? Man uulxgit, ada tgi duulxgit
ndaa dzi yaak-t man duulxk-it, ada tgi duulxk-it
where irr go-3.ii up stuck-3.ii and down stuck-3.ii
‘Where could he go? He could not go up, he could not go down.’
(Boas 1912, The Story of Asdiwaal; 144–5)

Shortly after the narrator poses these RQs, Asdiwaal dies on the mountain.

The examples in (27) and (28) show lines from narratives that have been translated with the negative existential, rather than the rhetorical content question translation.

(27) Narrative context: Sts’ool (Beaver) strands Awta (Porcupine) on an island in the lake. Awta wakes up after nearly drowning, he can’t swim:
Ada ndaa dza dzagayaakit gisga dmt goosga gilhawlitga
ada ndaa dzi dzaga-yaak-t gisga dm=t goo=sga gilhawli-t-ga
then how irr across-go-3.ii prep prosp=3.i go=cn ashore-3.ii=cn
‘And he had no way to go ashore’
(Boas 1912, Story of Porcupine and Beaver; 232–3)
(28) ’Nax’nuuyu amhawsm ada goo dzi hawyu.
’nax’nuu-i-u amhaw-sm ada goo dzi haw-u
hear-tr-1sg.ii voice-2pl.ii and what irr say-1sg. ii
‘I have heard your request and there is nothing I can say.’
(Beynon 1932–1939, When Txagaaxs Embarrassed Ligeex; Manuscript 91, 5)

Elicitations with fluent speakers confirm that marked RQs may variably be translated as content questions or negative assertions (29), and that this construction is associated with the negative implication: marked RQs only allow negative answers and are only felicitious in non-information seeking scenarios such as (30).

(29) Goo dzi wila waali… Nah baaltu txa’nii goo
Goo dzi wila waal-I nah baal-T-u txa’nii goo
what irr manr do-irr.1sg.ii pfv try-T-1sg.ii what
‘What could I do? I’ve tried everything.’
‘I don’t know what to do, I’ve tried everything.’
‘There’s nothing to do, I’ve tried everything.’
(30) Context: Bill and Lucy are driving to a cabin, and halfway there, they realize they forgot the cooler with all the food. Arriving at the cabin, Bill says “Kwdii”nu’ [I’m hungry]:
L: Gaba ligi goo!
gap=a ligi goo!
eat=cn ligi what
‘Eat something.’
B: Goo dza gabi (duu)! Akadi goo!
goo dzi gap-i duu aka=di goo
what irr eat-irr.1sg.ii eh neg=foc what
‘What’s there to eat, eh, nothing!’

Though (30) shows that the speaker may answer their own question, it is also possible for the addressee to answer it.

(31) Context: Same as (30):
B: Goo dza gabi (duu)?
goo dzi gap-I duu
what irr eat-irr.1sg.ii eh
‘What’s there to eat, eh?’
L: Akadi goo…
aka=di goo
neg=foc what

In the following, we see that marked RQs are not felicitous in information-seeking contexts in which there is a possible (non-negative) answer.

(32) Context: Allie and Ben just went shopping, the fridge is full of food. Allie says “Kwdii”nu’ [I’m hungry]:
B: Gaba ligi goo!
gab=a ligi goo
eat=cn ligi what
‘Eat something.’
A: # Goo dzi gabi, heelda goo doot dm gabm.
goo dzi gab-i, heelda goo doo-t dm gap-m.
what irr eat irr.1sg.ii much what have-3.ii prosp eat-1pl.ii
Intended: What’s there to eat? There’s so much to eat.

The following infelicitous sentence shows that, unlike the unmarked RQs described earlier, marked RQs cannot function as the “obvious positive answer” flavor described in Caponigro and Sprouse (2007).

(33) Context (adapted from Caponigro and Sprouse 2007): Al is worried Cassie didn’t have fun at his party. They both know that Cassie was dancing all night at the party. Bill says ‘Gooyu nahawn?’ [what do you mean?]:
# naa dzi nah miilg-it a ludaba aatk?
who irr pfv dance-sx prep whole night
Intended: ‘Who was dancing all night?’

Turning to embedding facts, we find that these dzi-marked questions may be embedded under rogative and responsive predicates – respectively, those that embed only interrogatives (such as “ask” and “wonder”), and those that embed interrogative and declarative complements (such as “know”).

(34) Nah güüdagu goo dzi wila gyooyi.
nah güüdax-i-u goo dzi wila gyoo-i
pfv ask-tr-1sg.ii what irr manr do-irr.1sg.ii
‘I asked what could I do.’
(35) Akandi wilaay goo dzi wila gyooyi.
aka=n=di wilaay goo dzi wila gyoo-i
neg=2sg.i=foc know what irr manr do-irr.1sg.ii
‘I don’t know what I could do.’

The following examples show that both a matrix and embedded marked RQ may be uttered in the same context.

(36) Context: You’re visiting a friend in a small town and realize you forgot your toothbrush, it’s Sunday night, and you know everything will be closed. Frustrated, you say to your host:
Oo du, ndeh dza gooyi gyaw’n?
oo du ndeh dzi goo-I gyaw’n
oh particle where irr go irr.1sg.ii now
‘Well, where could I go now?’
(37) Context: Same as (36):
Aamł güüdaguwii ndeh dza gooyi gyaw’n?
aam=ł güüdax-u=ii a=ł ndeh dzi goo-i gyaw’n ask-1sg.ii=q where irr go-irr.1sg.ii now
‘Should I even ask where I could go now?’

The ability for dzi-marked questions to be embedded under question-embedding verbs points to shared syntactic and/or semantic properties between this flavor of question and ISQs. I propose the following structure for marked RQs: as we saw in (25), a wh-expression undergoes movement to the specifier of CP; however, in this construction, the CP projection is headed by dzi. This explains how this construction does not allow the presence of the question particle u.

(38) Marked RQs

In sum, we find two strategies for forming RQs in Sm’algyax. The first kind (unmarked-RQs) is unremarkable. It resembles an ISQ in terms of its mor-phosyntax, but differs in terms of the discourse contexts it appears in. Marked RQs, on the other hand, differ from ISQs in terms of their morphosyntactic characteristics (the appearance of dzi, the absence of u), and they differ from unmarked RQs in that they are associated with a strict negative implication. There are, however, reasons to treat this construction as being an interrogative clause: it may be embedded under verbs that embed questions, and also bear other features associated with ISQs (for example, wh-movement). Available responses, such as answering, as well as agreeing or disagreeing, point to variability between more question-like and assertion-like behavior: it is not typical for an addressee to agree to a question, or to answer an assertion. In lieu of a full, compositional analysis, I outline in the following section some potential avenues for analyzing marked and unmarked RQs in Sm’algyax.

5 RQs as questions

One prominent approach to analyzing RQs treats them as being syntactically and semantically equivalent to ISQs, differing from ISQs in terms of their pragmatics (Biezma and Rawlins 2017, Caponigro and Sprouse 2007). Evidence for this isomorphism comes from answerability and embeddability tests.

Caponigro and Sprouse (2007) characterize both information-seeking and rhetorical interrogatives as denoting the proposition that is its true complete answer in a given world w.

(39) [RQ] W = [ISQ] W = p: p is the true complete answer to RQ or ISQ in w.

In terms of providing an analysis for the pragmatics of RQs, Caponigro and Sprouse suggest that a question Q is rhetorical if and only if the denotation of Q is a member of a set CG S−A , the set of all propositions believed by the speaker and the addressee (their “Common Ground”).

(40) CG S−A = {p: p is mutually believed by the Speaker and the Addressee}
(41) Q is a RQ iff [Q] W ∈ CG S−A

Adopting this style of approach is straightforward for unmarked RQs in Sm’algyax. The answerability facts showing that they are compatible with positive and negative answers, as well as the presence of the interrogative particle u that is restricted to root-level interrogatives, suggest that these are in fact questions that happen to be licensed in contexts distinct from those which license ordinary ISQs. This places unmarked RQs alongside the English RQs described in Caponigro and Sprouse (2007) as parsimoniously lending themselves to a “RQs as questions” analysis, contra the family of analyses that analyze RQs as negative assertions (e.g. Han 2002, Sadock 1974).

Accounting for irrealis-marked RQs is less straightforward. We observe with this flavor a negative inference, which some analyses of RQs have provided accounts for: Sadock (1974) and Han (2002) analyze RQs as covert negative assertions, while Ladusaw (1980) analyzes them as questions with an empty answer-set.

Han (2002) formalizes the questions as negative assertions by suggesting that as part of a post-LF derivation, the wh-expression maps onto a negative quantifier. For instance, a RQ “what does he (even) like” would be interpreted not as a question, but as the negative assertion “he likes nothing.” However, the embedding facts discussed in the previous section as well as the acceptability of answering marked-RQs suggest that we should treat these constructions as formally interrogative clauses.

A promising approach, then, is Ladusaw (1980), which treats RQs as questions with an empty answer-set, and that RQs may be worded in a way that facilitates the expected answer. Evidence for this approach in English RQs comes from the ability for minimizers, a kind of “strong” negative polarity item, such as budge an inch or lift a finger to be licensed in negative RQs, and not ISQs. We see an example of a minimizer appearing in the following question, triggering an obligatory negative RQ.

(42) Who lifted a finger to help? RQ

If we assume a domain with the individuals Allie and Ben, the possible answers for (42) may be interpreted as the following in (43). The only answer that licenses the minimizer, given in bold, is the negative answer.

(43) Allie and Ben lifted a finger to help.
Allie lifted a finger to help.
Ben lifted a finger to help.
Nobody lifted a finger to help.

(42) thus conveys the Speaker’s belief that nobody helped, because only the negative answer Nobody lifted a finger to help licenses the minimizer lift a finger.

Turning back to Sm’algyax, we observe in marked RQs a construction where the clause that typically corresponds to a presupposed clause in a canonical content question is introduced by dzi: a subordinator that has been characterized in §3.2 as an element that introduces uncertain or unreal information. Perhaps, then, the presence of dzi in marked RQs, like the presence of minimizers in English questions, signals to the addressee that the negative answer is expected.

One potential piece of evidence pointing to this analysis for dzi-marked questions comes from the optional appearance of dzi in declarative sentences with fronted negative quantifiers (which are composed of negation and a wh-expression). The presence of overt negation distinguishes these constructions from marked-RQs.

(44) Ada ałga goo dza niidzit.
ada ał=ga goo dzi niist-i-t
and neg=foc what irr see-tr-3.ii
‘And they didn’t see a thing.’

(Beynon 1932–1939, The myth of what happened when one of the princesses made a pet of a grubworm; Manuscript 129, 21)[4]

Assuming the same question–answer congruence as in (42)–(43) for Sm’algyax, it is only the negative existential answer that may co-occur with dzi.

(45) Goo dzi gabi?
goo dzi gap-i
what irr eat-irr.1sg.ii
‘What could I eat?’
(46) *Sami dił anaay dzi gabi [meat and bread [irr eat-I ]]
*Sami dzi gabi [meat [irr eat-I]]
*Anaay dzi gabi [bread [irr eat-I]]
Akadi goo dzi gabi [nothing [irr eat-I]]

Though more research needs to be done to adequately account for the meaning contribution of dzi appearing in and outside of questions, an analysis of RQs as questions with a negative answer is a promising avenue for analyzing marked RQs in Sm’algyax.

6 Conclusion

This article presented a first pass at describing and comparing two RQ constructions in Sm’algyax. The first is referred to as “unmarked” due to its resemblance to ISQs with respect to its morphosyntax. In terms of answerability, this type is compatible with positive and negative answers. “Marked” RQs feature irrealis subordinator dzi and the obligatory absence of the interrogative particle u. Beyond these differences, they function as obligatorily RQs, with an obligatory “negative implication.” I outlined some possible directions how to analyze (“negative”) RQs, and particles associated with the negative inference, suggesting that a Ladusaw-esque 1980 approach, in which the elements appearing within the question signaled the speaker’s attitude or bias toward a negative answer, fared better than approaches analyzing questions as negative assertions (Han 2002, Sadock 1974).

One contribution of this article that warrants further investigation is that Sm’algyax seems to possess a uniquely RQ construction, which is associated with a negative implication. We find a similar construction in Japanese mono ka RQs, which are also obligatorily negative RQs, analyzed as involving an irrealis clause and a covert negative operator (Oguro 2014, 2018).

(47) Dare-ga kur-u mono ka!
who-nom come-prs comp q
‘No one will come!

How common is this, crosslinguistically, for irrealis marking in interrogatives to be associated with a rhetorical effect? In languages beyond Sm’algyax, we also find irrealis marking occurring in RQs. For example, subjunctive/irrealis-marked interrogatives in Fula (Senegambian) are interpreted as RQs (Palmer 2001, 111). However, the presence of subjunctive/irrealis (without any other marker) may also be associated with forming canonical ISQs. This is suggested to be the case for Caddo (Caddoan), Hixkaryana (Cariban), and Serrano (Uto-Aztecan) (Palmer 2001, 172). Closer investigation of RQ formation across a diverse set of languages is needed to uncover the patterns and variation in the appearance of irrealis marking in interrogatives and its association with the formation of RQs.


Morpheme glosses


first person


second person


third person


agent extraction morpheme,




common noun connective








series I clitic


series II suffix


series III pronoun








manner subordinator












proper noun connective














question particle


reason subordinator






subject extraction morpheme



Additional abbreviations used




common ground


information-seeking question




rhetorical question






I would like to thank the Ts’msyen elders who have graciously shared their language with me. In particular, Velna Nelson, Beatrice Robinson, and Ellen Mason. I would also like to thank the attendees of the workshop Non-canonical interrogatives across languages: prosody, semantics, pragmatics as well as two anonymous reviewers for their valuable questions and comments.

  1. Funding information: This article draws on research supported by the Social Sciences and Humanities Research Council.

  2. Conflict of interest: Author states no conflict of interest.


Beynon, William. 1932–1939. “Tsimshian texts.” Manuscript. Special Collections. Columbia University Archives.Search in Google Scholar

Biezma, María and Kyle Rawlins. 2017. “Rhetorical questions: Severing questioning form asking.” In Proceedings of semantics and linguistic theory, vol. 27.10.3765/salt.v27i0.4155Search in Google Scholar

Boas, Franz. 1912. “Tsimshian texts (new series).” In Publications of the American ethnological society, vol. 3, Leyden, Late E.J. Brill.Search in Google Scholar

Braun, Bettina, Nicole Dehé, Jana Neitsch, Daniela Wochner, and Katharina Zahner. 2019. “The prosody of rhetorical and information-seeking questions in German.” Language and Speech 62(4), 779–807.10.1177/0023830918816351Search in Google Scholar

Brown, Colin. 2022. “Questions and their relatives in Sm’algyax.” Manuscript, in Google Scholar

Caponigro, Ivano and Jon Sprouse. 2007. “Rhetorical questions as questions.” In Proceedings of Sinn und Bedeutung, vol. 11, edited by Estela Puig-Waldmüller, p. 121–33.Search in Google Scholar

Dunn, John Asher. 1979. “A reference grammar for the Coast Tsimshian language.” Ottawa: National Museums of Canada.10.1353/book65525Search in Google Scholar

Evans, Nicholas. 2007. “Insubordination and its uses.” In Finiteness, edited by Irina Nikolaeva, Oxford: Oxford University Press.Search in Google Scholar

Farkas, Donka F. 2022. “Non-intrusive questions as a special type of non-canonical questions.” Journal of Semantics 39(2). 295–337. 10.1093/jos/ffac001.Search in Google Scholar

Forbes, Clarissa. 2023. “Tsimshianic.” In The languages and linguistics of Indigenous North America: A comprehensive guide, vol. 1, edited by Carmen Jany, Marianne Mithun, and Keren Rice, De Gruyter, Mouton.Search in Google Scholar

Han, Chung-hye. 2002. “Interpreting interrogatives as rhetorical questions.” Lingua 112(3), 201–29. 10.1016/S0024-3841(01)00044-4.Search in Google Scholar

Hunt, Katharine. 1993. “Clause structure, agreement and case in Gitksan.” PhD thesis. University of British Columbia.Search in Google Scholar

Ladusaw, William. 1980. “Polarity sensitivity as inherent scope relations.” PhD thesis. Austin: University of Texas.Search in Google Scholar

Matthewson, Lisa. 2004. “On the methodology of semantic fieldwork.” International Journal of American Linguistics 70, 369–415.10.1086/429207Search in Google Scholar

Mulder, Jean. 1994. “Ergativity in Coast Tsimshian (Sm’algyax).” PhD thesis. Berkeley: University of California.Search in Google Scholar

Oguro, Takeshi. 2014. “Negation in certain rhetorical questions in Japanese.” In Proceedings of the Florida Linguistics Yearly Meeting, vol. 1.Search in Google Scholar

Oguro, Takeshi. 2018. “Properties of mono ka rhetorical questions.” In Mit working papers in linguistics 88: Proceedings of the 13th Workshop on Altaic Formal Linguistics, vol. 13, edited by Céleste Guillemot, Tomoyuki Yoshida, and Seunghun J. Lee, p. 193–204.Search in Google Scholar

Palmer, Frank Robert. 2001. Mood and modality. Cambridge University Press reprint, revised edition. Bruce Rigsby. “Gitxsan grammar.” Manuscript, 1986. University of Queensland, Australia.Search in Google Scholar

Rigsby, Bruce. 1986. “Gitxsan grammar.” Manuscript. Australia: University of Queensland.Search in Google Scholar

Sadock, Jerold. 1974. Towards a linguistic theory of speech acts. New York: Academic Press.Search in Google Scholar

Sasama, Fumiko. 2001. “A descriptive study of the Coast Tsimshian morphology.” PhD thesis. Kyoto University. SLLTD. = “Sm’algyax Living Legacy Talking Dictionary. (19.05.2023).Search in Google Scholar

Tarpent, Marie-Lucie. 1987. “A grammar of the Nisgha language.” PhD thesis. University of Victoria.Search in Google Scholar

TFS Working Group. “Feeding fluffy.” Totem Field Storyboards. (19.05.2023).Search in Google Scholar

Received: 2022-11-16
Revised: 2023-05-22
Accepted: 2023-05-27
Published Online: 2023-07-19

© 2023 the author(s), published by De Gruyter

This work is licensed under the Creative Commons Attribution 4.0 International License.

Downloaded on 29.9.2023 from
Scroll to top button